COMP 551 MIDTERM 2 EXAM SAMPLE
QUESTIONS REVIEW (Applied Machine
Learning) McGill University
MCGILL UNIVERSITY
FALL SEMESTER, 2023
Campus: Montréal - Downtown
COMPUTER SCIENCE
Applied Machine Learning Course
NOTE: - You have 75 minutes to complete this exam.
- This examination contains 8 pages, including this page, and contains 21 questions
- Please turn off your phone during the exam, or put them airplane mode.
- Write your answers on the question booklet. Please try to avoid writing on the backs of pages. If
you must do so, please indicate clearly on the front of the page that you have written on the back of
the page.
- You may use your 5 double-sided hand-written notes for this exam (no printed notes allowed, even
if you handwrote them on ipad then printed them out). You may not use any other resources such
as printed lecture notes, books, or any online searches, etc. Calculators are allowed if they are non-
programmable, that is cannot store any text-notes or programmes which you wrote.
By writing my name below, I confirm that all my exam work will be done entirely by myself, with no
help from others.
McGill ID No: Student name:
, COMP 551 - Midterm
Section Type Points
A True/False 10
B Short Answers 10
C Derivations 45
D Scenarios 15
Total 80
SECTION A: True/False Questions
Simply indicate True or False for each statement.
1. When a model has a higher variance, over fitting is more likely. Is this true or false? [1 point]
2. When doing mini-batch stochastic gradient descent with mini-batch size 5, one can always
reuse the same 5 data points to calculate the gradient. Is this true or false? [1 point]